AAAI-96 Tutorial on

Designing Computational Markets and Multiagent Organizations

Michael Wellman and Tad Hogg

Delivered in Portland, OR on 5 August 1996

Description


The recent explosion of internet activity and development of software agents heralds a time when autonomous computational processes on wide-area networks will be deployed on behalf of human users. Given their varying goals, capabilities, and resources, computational agents will often find it necessary to coordinate their activities to achieve desired results. The problem facing designers of agents and interaction protocols is how to achieve an allocation of activities and resources that best meets overall objectives, without imposing centralized control. More generally, how can we relate the global behavior of a collection of agents to the local behavior of individuals? Given a large number of autonomous agents, each working with a limited view of the overall situation and perhaps with conflicting goals, under what conditions can we expect to produce good solutions to complex problems?

In this tutorial we address the fundamental problem of coordinating multiple agents through the use of market mechanisms and organizational structures. We present some relevant background in economics and organization theory necessary to understand these systems, leading to some general design methodology for constructing computational economies and organizations. The methods are elaborated through case studies (e.g., networked information services), computational experience, and discussion of key technical issues (e.g., dynamic behavior) underlying multiagent systems.

Presenters


Michael Wellman is an Assistant Professor in the Department of Electrical Engineering and Computer Science at the University of Michigan. He received a PhD in Computer Science from the Massachusetts Institute of Technology in 1988 for his work in qualitative probabilistic reasoning and decision-theoretic planning. Current research focuses on computational market mechanisms for distributed decision making. In 1994, he received an NSF National Young Investigator award.

Tad Hogg is a member of the research staff at the Xerox Palo Alto Research Center. His research interests include dynamics of multiagent systems, the use of economic mechanisms for resource allocation, cooperative problem solving and analogies with physical phase transitions found in combinatorial search problems. He holds a PhD in physics from Stanford University.

Outline

Bibliography

updated 14 Aug 96